Block(0) - stacking context
	Block(1) - Default hit flags
		Inline(2) - Default hit flags
	Block(3) - Something weird, maybe different flags or agr?
		Inline(4)

Backgrounds for Block(1) and Block(3) should both go into the list for step 4 (BlockBorderBackgrounds() in gecko), and the backgrounds for Inline(2) and Inline(4) will go into the list for step 7 (Content() in gecko [1]) as part of 7.2.1.1

So the DL order is:

nsDisplayBackgroundColor(Block(1))
nsDisplayBackgroundColor(Block(3))
nsDisplayBackgroundColor(Inline(2))
nsDisplayBackgroundColor(Inline(4))


If we use the parent optimization to add hit test info, then we'd skip adding hit test info for Inline(2), since Block(1) already added it. This is incorrect though, since Block(3) will also add hit-test info (because of its weird flags), and it'll be behind Inline(2).

Trying to scroll with the mouse over Inline(2) will give the behaviour for Block(3), even though that's not what we want.



[1] Backgrounds (and hit test info) always get added to the BorderBackgrounds() list, but the code below creates a temporary nsDisplayListCollection where the BorderBackgrounds() list is a pointer to the outer collection's Content() list.

https://searchfox.org/mozilla-central/rev/a7315d78417179b151fef6108f2bce14786ba64d/layout/generic/nsBlockFrame.cpp#6344